Bubble Production Mechanism in a Microfluidic Foam Generator

We present the design and characterization of a microfluidic bubble generator that has the potential of producing monodisperse bubbles in 256 production channels that can operate in parallel. For a single production channel we demonstrate a production rate of up to 4 kHz with a coefficient of variation of less than 1%. We observe a two-stage bubble production mechanism: initially the gas spreads onto a shallow terrace, and then overflows into a larger foam collection channel; pinning of the liquid-gas meniscus is observed at the terrace edge, the result being an asymmetric pinch-off. A semiempirical physical model predicts the scaling of bubble size with fluid viscosity and gas pressure from measurements of the pinned meniscus width.

Figure 1

Schematic of the device geometry (not to scale), showing three consecutive production channels. The glass wafer that seals the device at the top was not represented.

Figure 2

(Top) Production cycle of a bubble (using the $2.5\mu \mathrm{m}$ device, a solution of 60% glycerol, gas pressure $P=0.5\mathrm{bar}$, and liquid flow rate $\varphi =50\mu \mathrm{l}/\min $). White dashed circles are guides to the eye. (Bottom) Variation of the curvature radius $R$ as a function of real time (a), and of time scaled using the production period (b), for the three glycerol solutions used.

Figure 3

Schematic of the pinning process at the terrace edge [(a) lateral view; (b) top view]. (c) The length of the pinned meniscus just before pinch-off ($L_{\mathrm{pinch}\mathrm{\text{−}}\mathrm{off}}$) as a function of gas pressure for the three solutions used.

Figure 4

The dependence of the “pancake spreading” process on terrace width, observed in the variable geometry device—here three different widths are shown, (a) $14\mu \mathrm{m}$, (b) $12\mu \mathrm{m}$, (c) $8\mu \mathrm{m}$, under otherwise identical experimental conditions. The shorter the terrace, the less the pancake spreads, resulting in a smaller bubble. In (a), the terrace width is too large to allow a bubble to form, resulting in complete spreading.

Figure 5

Experimentally measured bubble volumes $V_b$ (a) for the $2.0\mu \mathrm{m}$ device, and the volumes scaled by viscosity of the liquid $V_b/\eta$ (b). In (c),(d) we show the predictions of the phenomenological model vs experimental data (the model data is based on the measurements of pinned meniscus length from Fig. 3c and assumes $\alpha =60$).